Delft University of Technology
Hydraulic-mechanical properties of microfaults in granitic rock using the Punch-Through
Shear test
Kluge, Christian; Blöcher, Guido; Barnhoorn, Auke; Bruhn, David
DOI
10.1016/j.ijrmms.2020.104393
Publication date
2020
Document Version
Final published version
Published in
International Journal of Rock Mechanics and Mining Sciences
Citation (APA)
Kluge, C., Blöcher, G., Barnhoorn, A., & Bruhn, D. (2020). Hydraulic-mechanical properties of microfaults in
granitic rock using the Punch-Through Shear test. International Journal of Rock Mechanics and Mining
Sciences, 134, 1-16. [104393]. https://doi.org/10.1016/j.ijrmms.2020.104393
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy
Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
International Journal of Rock Mechanics & Mining Sciences 134 (2020) 104393
Available online 22 August 2020
1365-1609/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Hydraulic-mechanical properties of microfaults in granitic rock using the
Punch-Through Shear test
Christian Kluge
a,b,*, Guido Bl¨ocher
a,c, Auke Barnhoorn
b, David Bruhn
a,baHelmholtz Centre Potsdam - GFZ German Research Centre for Geosciences, Department of Geoenergy, Telegrafenberg, 14473, Potsdam, Germany bDepartment of Geoscience and Engineering, Delft University of Technology, PO-box 5048, 2600 GA, Delft, the Netherlands
cInstitut de Physique du Globe de Strasbourg, UMR 7516, 5 rue Ren´e Descartes, 67084, Strasbourg Cedex, France
A R T I C L E I N F O Keywords: Shear fracture Microfault Permeability Aperture Sustainability Granite A B S T R A C T
Fault zones are key features in crystalline geothermal reservoirs or in other subsurface environments due to the fact that they act as main fluid pathways. An adequate experimental description of the evolution of permeability of a realistic microscopic fault zone under in-situ reservoir and fracture parallel flow conditions is required. To address this topic, we demonstrate a novel experimental set up (Punch-Through Shear test) that is able to generate a realistic shear zone (microfault) under in-situ reservoir conditions while simultaneously measuring permeability and dilation. Three samples of intact granite from the Odenwald (Upper Rhine Graben) were placed into a MTS 815 tri-axial compression cell, where a self-designed piston assembly punched down the inner cyl-inder of the sample creating the desired microfault geometry with a given offset. Permeability was measured and fracture dilation was inferred from an LVDT extensometer chain, as well as the balance of fluid volume flowing in and out of the sample. After fracture generation, the shear displacement was increased to 1.2 mm and pore pressure changes of ± 5 or ± 10 MPa were applied cyclically to simulate injection and production scenarios. Formation of a microfault increased the permeability of the granite rock by 2 to almost 3 orders of magnitude. Further shear displacement led to a small increase in permeability by a factor of 1.1 to 4.0, but permeability was reduced by a factor of 2.5 to 4 within 16 h due to compaction and fault healing. Effective pressure cycling led to reversible permeability changes. CT images showed that the fracture network is rather complex, but depicts all features commonly observed in larger scale fault zones.
1. Introduction
The permeability of fractures and faults, or, more generally, shear fractures in crustal rocks has been a substantial research topic in the past. It is not only important in crustal faulting processes or earthquake mechanisms, but is also key in understanding the fluid flow in faulted or enhanced geothermal systems,1–4 as well as the stability of underground
constructions, such as tunnels or nuclear waste repositories. Fractures control the hydrological and mechanical behaviour of rock masses, such that any changes in fracture properties have a large impact on the bulk transport properties. Laboratory experiments with fractured rock sam-ples are a substantial element to better understand and physically characterise fractures in order to imply physical relationships for the larger scale.
Conventional tri-axial compression tests are classically used to
generate shear fractures or shear zones. Since measuring fracture permeability in such experiments is difficult and fracture permeability cannot be isolated from the rock matrix, other experimental approaches are required. The fractures used for experiments are commonly gener-ated in tensile mode, for example by Brazilian Disk splitting,5–7 or
existing, as well as artificial joints are used.8–10 The sample halves are
then placed together at a certain offset and are installed in a tri-axial cell or shear-box tool to study the permeability evolution. Although the term shear fracture is often applied to such geometries, they do not represent the complex structure of a single shear or fault zone. Several attempts have been made in the past to reproduce a single realistic fault zone in the laboratory. Watanabe et al.11 used a shear-box tool to apply mode II
loading to an intact rock sample, but pointed out that it is difficult to generate a single shear fracture from shear-box experiments due to the occurrence of multiple fractures. However, experiments by Frash et al.12
* Corresponding author. Helmholtz Centre Potsdam - GFZ German Research Centre for Geosciences, Department of Geoenergy, Telegrafenberg, 14473, Potsdam, Germany.
E-mail address: christian.kluge@gfz-potsdam.de (C. Kluge).
Contents lists available at ScienceDirect
International Journal of Rock Mechanics and Mining Sciences
journal homepage: http://www.elsevier.com/locate/ijrmms
https://doi.org/10.1016/j.ijrmms.2020.104393
using shale and anhydrite samples highlighted the importance of fea-tures, such as en echelon strucfea-tures, for the aperture anisotropy. Recently, Ye & Ghassemi13 used a conventional tri-axial setup with
cy-lindrical double-flawed samples connected to pumps using two injection holes. By increasing the injection pressure, they generated a complex fracture network including shear and tensile fractures and compared the permeability before and after fracture generation.
Besides the challenge to generate a single shear zone while flowing fluid through the sample, permeability is rarely measured during the initiation and propagation of a fracture under in-situ pressure condi-tions. First attempts to measure permeability while generating a shear fracture in high permeable rocks have been made for example by Zhu et al.14 or Crawford.15 De Paola et al.16 measured the permeability of
anhydrite samples during tri-axial testing and subdivided the perme-ability evolution into three stages of compaction, dilation and brittle failure. Others measured the permeability in tri-axial tests using the pore pressure oscillation method.17–19 Mitchell and Faulkner20 quantified the
permeability enhancement to about 2 orders of magnitude with the same experimental procedure for granite and granodiorite. However, the shear fractures in these tests were not parallel to the flow direction or the entire sample integrity was lost during failure, which makes it difficult to isolate the mechanical and hydraulic behaviour of the frac-ture from the surrounding rock mass. Watanabe et al.11 fractured granite
samples in a shear-box tool and measured the permeability of these samples in another flow-through apparatus. Although a flow parallel fracture geometry is given, no fluid flow was applied during fracturing and the sample had to be removed from the shear apparatus after frac-ture generation, which highly alters the hydraulic-mechanical fracfrac-ture properties.
The permeability during shearing is substantial for the understand-ing of the processes involved in hydro-shearunderstand-ing or the permeability evolution in active faults. In some experimental studies, asymmetric loading conditions on a pre-existing tensile fracture in cylindrical sam-ples with different end cap designs were used to displace a fracture while measuring permeability.17,21,22 Shear-box tools were also used to
displace a fracture under normal loading conditions,23,24 but the normal
effective pressures in these tests were too low to make meaningful conclusions or additional fractures formed. Watanabe et al.11 compared
the permeability of displaced tensile fractures and fractures generated by shearing and found that displaced tensile fractures have a higher permeability and a higher resistance to closing compared to shear fractures when increasing normal stress. More frequently, tri-axial set ups with injection-holes are used to generate a shear displacement on a pre-existing planar shear plane in different materials.25,26 Other authors
performed performed stick-slip experiments with artificial surfaces,27,28
which are sheared while measuring permeability. Although it is possible to accurately quantify permeability, the experimental set ups mentioned involve unrealistic planar fracture surfaces, which lead to only marginal permeability changes or are performed at very low normal stress.
The longevity or sustainability of the generated permeability enhancement by fracturing is crucial for engineering applications, in which, for example, pore pressures change due to production and in-jection schemes, or the fracture aperture changes due to stress relaxation when the well is not operated. In experimental studies with granite5,11,29,30 the effective pressure was varied by changing the
confining pressure cyclically using displaced tensile fractures in granite and found a progressive reduction in permeability with each cycle. However, manually displaced tensile fractures might have a different initial strength during the first loading stages compared to faults. Furthermore, a variation in confining pressure is different compared to changing the pore pressure when considering effective pressure co-efficients.31 Zangerl et al.32 evaluated the fracture normal stiffness of
several granite samples and provided a range of values, without finding systematic changes, but highlighted the importance of surface geometry and asperity deformability. Lastly, permeability losses are often observed in experiments and are related to mechanical compaction17,27
or fluid-rock interactions.33 Therefore, effective pressure cycling,
frac-ture stiffness characterisation, as well as compaction induced perme-ability reductions need to be considered for complex shear fracture or fault geometries. There is too little quantification of permeability enhancement in granite during shear fracture generation or perme-ability evolution during shearing and only a few of those studies account for a realistic shear zone geometry, or they are limited in terms of the displacement magnitude.
We present an approach to combine most of the above-mentioned experimental technologies in a single experiment. The evolution of permeability during initiation and propagation of a circular shear frac-ture (referred to as a microfault in the following) is presented, using the Punch-Through-Shear (PTS) test, originally developed to measure mode II fracture toughness.34–38 The existing experimental setup has been supplemented with the possibility of saturating the sample and allowing fluid flow-through to measure the pressure gradient across the sample. This enables us to determine the permeability of an artificially-generated microfault with flow oriented parallel to the fracture. The PTS test allows for the quantification of the permeability during fracture initiation, propagation and effective pressure changes, an analysis of fracture properties and geometry, as well as an assessment of its hydraulic-mechanical sustainability.
2. Materials & methods 2.1. Testing equipment
The Punch-Through-Shear (PTS) test was carried out in a conven-tional MTS tri-axial compression cell. The stiff, servo-controlled loading frame (MTS 815, Material Testing Systems Corporation) holds a loading capacity of up to 4600 kN (load cell calibrated to 1000 kN, calibration error < 1%) and a servo-controlled maximum hydrostatic confining pressure of 140 MPa applied via an oil-filled pressure vessel coupled to an external pressure intensifier. The pore fluid pressure was applied via four Quizix fluid pressure pumps (Model C6000-10K-HC-AT) with a maximum fluid pressure of 70 MPa. Flow-through was continuously applied at a minimum of 2 MPa confining pressure using two paired upstream pumps and two paired downstream pumps. The differential fluid pressure, which is the difference between in- and outflow pressure, was measured using a differential pressure sensor (Honeywell HL-Z; range: 1 MPa; line pressure max. 35 MPa; precision: ~1%). Changes in circumferential strain were measured using a LVDT extensometer chain. All experiments have been performed at ambient conditions, i.e. temperatures of 25–30 ◦C. Data were recorded at a frequency of 1 Hz. A
detailed description of the machine is found in Pei et al.39 2.2. Sample material
The testing material is a granite from the Odenwald region (International Geo Sample Number (IGSN): GFTRE0033) taken from a quarry in Rimbach, Germany. The fresh granite shows no signs of strong alterations or microfracturing (Fig. 1c–f). It is composed of quartz (Qtz) with a grain size ranging from 1 to 3 mm, feldspar (Fsp) between 2 to 3 mm and mica (Mca) between 1 to 3 mm, suggesting a very even size distribution of minerals. Quartz, as well as the darker and brighter mica are the main constituents, feldspars have a less frequent occurrence. Microfractures are only partially present and either cut through grains, or propagate along grain boundaries with no visible shear offset. The porosity of less than 0.6% is mostly intra-granular and rarely connected. The intact rock properties of the Odenwald granite, in the following labelled as PGR6-RI,40 are summarized in Table 1. Three granite samples
have been tested with their testing conditions and sample dimensions listed in Table 2.
International Journal of Rock Mechanics and Mining Sciences 134 (2020) 104393
2.3. Sample geometry & experimental setup
The idea of the Punch-Through-Shear (PTS) test is to have a notched intact cylindrical sample, of which a smaller inner part (inner cylinder) is punched through the surrounding hollow cylinder. The main objective is to generate a shear stress concentration at the notch tips generating a circular shear fracture (microfault) connecting the bottom and top notch of a cylindrical rock sample. The experimental set up and sample ge-ometry are illustrated in Fig. 1a and b.
The set up consists of a piston assembly that applies a load to the top inner cylinder and the bottom outer annulus of the sample, as well as a core-holder system supporting the sample at the bottom, both made of stainless steel. The cylindrical samples were drilled from larger blocks to a diameter, d, of 50 mm using a diamond drill bit and water as lubricant. The samples were cut to length, L, of 50 mm using a diamond saw, and the end-surfaces were ground plan-parallel. The notches were then drilled using a diamond drill bit with an inner diameter, rID, of 25 mm
and a wall thickness, i.e. notch width, WN, of 1 mm. Top notch depth and
lower notch depth were 5 and 15 mm respectively, resulting in an intact portion, LIP, of 30 mm. This intact portion, LIP, was larger compared to
the 15 mm suggested by the ISRM International Society for Rock Me-chanics and Rock Engineering,35 but was used to enable the proper
characterisation of the hydraulic, mechanical and geometrical proper-ties of the microfault. This deviation from the recommended design is acceptable, as no variations in fracture toughness up to a length of 35 mm were previously found.35,37,38 The mechanical integrity of the
top-notch walls was supported by a 0.3 mm strong steel ring (2 mm shorter than the top notch length LUN) to prevent breakouts and
compaction that were observed in pre-tests without a supporting ring (Fig. 1d). Due to its shorter length the ring is assumed to not cause axial stress to the rock sample itself. The sample was covered with a heat-shrink tube to prevent confining fluid from entering the sample.
The loading assembly, shown in Fig. 1a, consists of a mobile steel cylinder of 25 mm in diameter (I), guided by a surrounding hollow
Fig. 1. a) Experimental set up and dimensions, b) Experimental set up, c) Odenwald granite sample (PGR6-RI), d) integrity test, e) and f) electron probe
micro-analyzer (EPMA) images of the intact rock.
Table 1
Intact rock properties of samples PGR6-RI according to Bl¨ocher et al.40
TS [MPa] KIC [MPa*m0.5] UCS [MPa] E [GPa] ν [− ] φ [%] kini [m2]
IGSN: GFTRE0033 11.8 1.347 120–142 39.9 to 47.6 0.19 to 0.26 >0.5 >10−18
TS: tensile strength, KIC: mode I fracture toughness, UCS: unixaixial compressive strength, E: Young’s modulus, v: Poisson ratio, φ: porosity, kini: initial permeability.
Table 2
List of samples, sample dimensions and testing conditions.
Sample ID Testing conditions LIP [mm], L [mm], d [mm] Additional measurements Comments
PGR6-RI-01-07 Failure: Pc =40 MPa, Pp =20 MPa; Pressure cycling: none
31.40, 49.4, 50.05 Integrity of electrical circuit of
silver paste, no LVDT extensometer chain Small leak at the outlet, no reliable fluid volume data without impact on permeability PGR6-RI-01-08 Failure:
Pc =40 MPa, Pp =20 MPa; Pressure cycling: 5 MPa
30.15, 50.15, 50.10 LVDT extensometer chain, x-ray CT scanning
PGR6-RI-01-09 Failure: Pc =40 MPa, Pp =20 MPa;
Pressure cycling: 10 MPa
30.18, 50.19, 50.10 LVDT extensometer chain, x-ray CT scanning
Pc: confining pressure, Pp: pore pressure, LIP: length of intact portion, L: sample length, d: sample diameter.
cylinder (II), which is 2 mm shorter than the mobile inner cylinder. The space in between the two parts is sealed by O-rings to prevent oil from entering the sample. The lower part of the assembly consists of a rigid core-holder system with a hollow cylindrical shape to counteract the applied load from the top (III). A movable smaller core-holder in the middle (IV), also sealed with O-rings, is mounted on a set of springs in an air-filled chamber. This is to prevent the inner cylinder from falling down instantly after failure and allows for a controlled displacement along the fracture. Fluid ports were incorporated within the loading assembly and core-holder system, sealed from the confining fluid by O- rings. A grid across the surface of the end caps allows for uniform in- and outflow across the entire cross-sectional area (top and bottom) of the sample. An extensometer chain (LVDT) attached at the centre between upper and lower notch is used to record the bulk circumferential strain throughout the entire duration of the test. Furthermore, the integrity of the sample wall during one experiment was monitored (PGR6-RI-01-07). Silver paste was applied to the circumference of the sample connected to an electronic circuit, such that any fracturing of the wall would result in an interruption of the electric signal (Fig. 1d).
2.4. Experimental procedures
The sample was installed in the measuring assembly and an absolute pore pressure of 10 mbar was generated using a vacuum pump (Laboxact SEM 820). The confining pressure was increased from 0 to 2 MPa at a rate of 0.5 MPa/min with the dry sample. After applying the vacuum, the system was saturated by applying a constant fluid pressure of 0.2 MPa from the upstream side. That way, water flew into the sample under nearly vacuum conditions, such that all pores were saturated. Saturation was finished when no more water flew into the sample to avoid effects of the saturation process to govern the permeability at later stages of the experiments.5 For all permeability measurements, a constant flow rate
(0.4–1 ml/min) was applied to the bottom of the sample and a constant pressure was applied at the downstream side (10–30 MPa) to determine the sample permeability. Steady-state conditions were reached when the flow rates and pressure difference reach a constant value. Sample permeability, k, was calculated by equation (1) using Darcy’s law,41
where the length, LIP, is the distance between the lower and upper notch
which corresponds to the maximum pressure gradient in the microfault. The sample permeability, k, corresponds to the permeability of the rock portion between the notches, assuming there is no pressure gradient along the 1 mm wide notches:
k = (Q / A)*(μLIP) /dP (1)
where Q is the volumetric inflow rate in m3/s, A is the cross-sectional
area of the sample in m,2 μ is the dynamic viscosity of the fluid (Pa s), LIP is the distance between the notches of the sample in m and dP is the differential fluid pressure in Pa. The minimum measurable permeability of our apparatus is around k = 10−18 m.2 It is important to note, that
steady-state permeability, kss, was measured during the hold phase (C),
during shearing (D) and before pressure cycling (E), as well as during the constant pressure phases during cyclic loading (F). In contrast, flow conditions were transient, ktrans, during the fracturing process (B) and
during pore pressure ramping (F). These measurements were labelled as apparent permeability in the results, although they were calculated using equation (1). Comparing the minimum measurable permeability and the permeability after the hold phase, the fold of increase in permeability, kFOI =ksshold/10−18, was calculated to infer the
perme-ability enhancement by fracturing.
During the fracturing process, an elevated effective pressure state is desired, where Terzaghi’s effective pressure,42 Peff (equation (2.1)), is
defined as the confining pressure, Pc, minus fluid pressure, Pp. The pore
fluid pressure was estimated by the outflow pressure, Pp,out, and the
differential fluid pressure, Pp,in–Pp,out, divided by two (equation (2.2)), assuming a linear pressure distribution5,6:
Peff=Pc− Pp (2.1) Pp=Pp,out+ ( Pp,in− Pp,out )/ 2 (2.2)
The change of volume of fluid in the sample, Vb (bulk volume) was
determined according to equation (3) from the difference of cumulative inflow (Vin) and cumulative outflow volume (Vout) in ml:
Vb=Vin− Vout (3)
After saturating the sample at 2 MPa confining pressure, the fluid pressure and confining pressure were simultaneously increased to 20 MPa and 40 MPa, respectively, within 1 h to an effective pressure of 20 MPa (Fig. 2, A). Afterwards, an axial displacement rate of 0.001 mm/s was applied to the top of sample (Fig. 2, B), which pushed the inner loading cylinder down relative to the guiding hollow cylinder until the point of failure. A drop in axial load, F, indicated the formation of a fracture. Up to this point, the inflow and outflow pressure were kept constant at a differential fluid pressure of 0.5 MPa. The fracture geom-etry was intended to initiate from the outer lower notch tip to the inner upper notch tip,35 resulting in an overall conical shape.
The mode II fracture toughness, KIIc, or critical stress intensity factor,
is a material parameter obtained by the PTS test, and depends on the type of rock material and its physical boundary conditions, such as confining pressure and temperature.36,37 The fracture toughness (K
IIc)
was calculated at the point of failure by the following empirical relation (equation (4)) based on the ISRM Suggested Methods35:
KIIc=7.74*10−2Fmax− 1.80*10−3Peff*10−3 (4)
where Fmax is the peak load in kN and Peff is the effective pressure in
MPa. Using the circumferential extensometer, the mechanical fracture dilation (emech) can be calculated by converting the change in
circum-ference (dU) in mm to a change in radius:
emech=dU/2π (5)
The fracture dilation is assumed to approximate the fracture aperture during the test. In addition, the changes in volume (equation (3)) were used to calculate the volume balance-based dilation evol. This was only
possible during phase B, since the change of cumulative fluid volume could be measured when a constant differential fluid pressure was applied. During fracture generation, additional void is generated. This is composed by the opening of the fracture itself, as well as the widening of the notches. The widening of the two notches is considered at the outer notch radius, rON. The additional volume can then be approximated as
follows:
ΔVnotch=Δe2rONπLN (6.1)
where rON is the radius to the outer notch, LN is the total length of the
upper and lower notch and e is the dilation with all units in m. The fracture has a shape of a truncated cone with a mantle surface area of (rON+rID)πLfrac. Here, Lfrac is the fracture length measured from the
lower outer and top inner notch, rID is the inner radius and rON the radius
to the outer notch wall (Fig. 6c). The additional volume due to fracture generation can then be calculated as follows:
ΔVfrac=Δe(rON+rID)πLfrac (6.2)
The total volume change measured ΔVmeasured =ΔVnotch+ΔVfrac, can
be used to estimate the fracture dilation, evol:
evol=Vbulk
/ ( (
2rONπLN+ (rON+rID)πLfrac
)
(6.3) After failure, the axial displacement ramp was stopped (hold phase) and the constant pressure boundaries for the fluid flow were changed to a constant inflow rate and a constant downstream pressure until a constant differential pressure of at least 0.1 MPa was reached and permeability was measured (Fig. 2, C). After that, the axial displacement
International Journal of Rock Mechanics and Mining Sciences 134 (2020) 104393
was again increased at a slower rate of 0.0001 mm/s leading to an in-crease in shear displacement until a maximum of around 1.2 mm was reached (Fig. 2, D). Finally, the axial displacement of the loading cyl-inder at the top was maintained to measure changes in stress and was kept in this position for at least 16 h (Fig. 2, E).
After that, the effective pressure was varied cyclically by changing the mean fluid pressure in the sample by ± 5 or ± 10 MPa (Fig. 2, F). A constant ramp operation of fluid pressure (0.5 MPa/min) was applied to the outflow side of the sample with the inflow pressure following the given ambient pressure change without active operation. Steady state flow conditions were achieved when the inflow and outflow pressure and rates reached a constant value, which took about 10 min consid-ering a hydraulic diffusivity of about 4.5*10-6 m2/s for a length of 0.05
m calculated according to Nicholas et al.43 For the diffusivity we used a conservative matrix permeability of 1*10-19 m2, Biot and Skempton
coefficients of 0.4 and 0.6 and a bulk compressibility of 30 GPa reported for granite.44 During pressure cycling, the vertical load, F, was trans-formed into shear stress,35 τ using equation (7):
τ=F/(πrIDLfrac
)
(7) where rID is the inner diameter in m and Lfrac is the fracture length in m.
This equation assumes and simplifies that the fracture to be a single plane. The fracture stiffness describes the fracture closure behaviour under increasing effective normal pressure. For a semi-logarithmic closure law45 only one free parameter, the fracture stiffness
character-istic, χ, is required and was obtained from the slope of the effective
normal pressure versus fracture aperture change32,46 following equation
(8): ln(Peff ) =χΔemech+ln ( σref n ) (8)
Peff is the normal effective pressure in MPa, χ, the fracture stiffness characteristic in mm−1, Δe
mech is the change in mechanical dilation in
mm and σrefn is any arbitrary reference value of normal stress in MPa.
After effective pressure cycling was finished, the testing conditions were kept constant for another 12 h, until the axial displacement, fluid and confining pressure were decreased simultaneously at similar rates as at the loading stage (Fig. 2, G).
3. Experimental results
In the following, the process of microfault generation and propaga-tion is described and the results for permeability evolupropaga-tion, flow rate changes, fracture dilation and volume balance during loading are given (Figs. 3–5). The evolution of shear stress, dilation, permeability and fracture stiffness for cyclic effective pressure loading are also presented (Figs. 7–9). Furthermore, the fracture geometry is described using the computed x-ray CT imaging (Fig. 10). The overall results are summa-rized in Table 3.
3.1. Fracture generation
The loading and failure process can be subdivided into five regimes with different dominant processes (Figs. 3–5). The stages of fracture initiation and propagation are illustrated in Fig. 6.
During stiffening and compaction at the first stage of loading, the volume of the sample was reduced and water was squeezed out of the sample. Between 0.60 to 0.65 mm of axial displacement, the inflow rate started to increase, the load curve flattened and the volume balance reached an inflection point (Fig. 3a and b; 4a, b, c and 5a, b, c). At this point, fracture initiation took place from the bottom of the sample, indicated by an increased inflow rate and a misbalance between in- and outflow, leading to dilation and a consequent increase of fracture vol-ume of the sample (Fig. 6). Using equations (5) and (6.3), the mechan-ical fracture dilation and the volume balance-based dilation was calculated, respectively, since it was assumed that a fracture is intro-duced into the sample. The mechanical dilation, which represents the change in mechanical aperture, increased faster than the volume-based
Fig. 2. Experimental procedure and flow chart of the respective phases (A–G).
dilation, since it took longer for the pore fluid to progress through the sample and the circumferential strain was only measured at the centre of the intact portion.
At the point of failure, the axial load reached values of about 212, 203 and 197 kN which is equal to 424, 406 and 394 MPa axial stress, which corresponds to 384, 366 and 354 MPa differential stress for sample PGR6-RI-01-07, PGR6-RI-01-08 and PGR6-RI-01-09, respec-tively. The load then dropped and the increase in outflow rate indicated a hydraulic short-cut between the lower and upper notch, suggesting a fully developed fracture system. Calculating the mean fracture propa-gation velocity, vprop, which is simply the length of the fracture, Lfrac,
over the time span of fracture initiation to the point of failure, tprop, the
velocity was in the range of 0.1–0.3 mm/s (vprop =Lfrac/tprop). After
failure, at a displacement of about 0.7–0.8 mm, the inner cylinder was punched down and most of the aperture was created, indicated by a rapid increase in mechanical and volume-based dilation. The measured load was now controlled by the frictional properties of the fracture. Shortly after the point of failure the displacement was put on hold for about 30 min (hold phase) such that the fluid pressure field could reach steady state flow conditions. The final value for dilation and fracture volume were reached during the hold phase when fluid pressure diffu-sion was finished. After restarting the displacement up to a maximum of 1.2 mm all samples showed no drop or rapid increase in axial load, i.e. no “stick slip” behaviour, but rather approached a limiting value of about 75 and 100 kN. This behaviour of fracture generation and prop-agation was similar in all three experiments. The fracture toughness values (equation (4)) for the saturated granite samples ranged between 15.3 and 16.4 MPa*m0.5 (Table 3).
3.2. Permeability and dilation evolution during failure
During all tests, the permeability was measured simultaneously during loading and failure. The flow conditions during failure were regarded as transient (apparent permeability, ktrans), but steady state
after the hold phase (steady-state permeability, kss). The permeability
evolution during fracture generation is shown in Figs. 3b, 4d and 5d,
where all experiments showed a pre-failure increase in apparent permeability of half to almost one order of magnitude, coinciding roughly with the inflection point of the volume balance as well as with the onset of yielding. For all granite samples, the increase in perme-ability was about 2 to almost 3 orders of magnitude at an effective pressure of 20 MPa (Pc =40 MPa, Pp =20 MPa), comparing the pre- failure permeability and the permeability during the hold phase (kFOI, Table 3). As explained before, the main dilation and consequent permeability was generated after the point of failure, with values of 0.06–0.07 mm (PGR6-RI-01-08) and 0.07–0.08 mm (PGR6-RI-01-09), while mechanical dilation and volume-based dilation were in very good agreement, suggesting that the measured dilation corresponds to the mechanical aperture. After the hold phase, the microfault was displaced up to 1.2 mm, resulting in a gentle increase in mechanical dilation to about 0.1 mm for sample PGR6-RI-01-08 and 0.09 mm in sample PGR6- RI-01-09, while permeability was consequently relatively constant. The maximum additional increase was between a factor of 4.0 for sample PGR6-RI-01-07 and 1.1 for sample PGR6-RI-01-09. No additional frac-turing was inferred from the mechanical-hydraulic and electrical data (PGR6-RI-01-07) during this period. Permeability decay curves measured for about 16 h were similar for two granite samples at 20 MPa effective pressure and showed a reduction by a factor of 2.5 to 4 (inset
Figs. 4d and 5d).
3.3. Permeability evolution during pore pressure cycling
To simulate reservoir operation procedures, such as injection or production, the pore pressure was cycled in two samples in order to monitor possible changes in shear stress due to slipping of the fault, permeability and strain at varying effective pressure (Figs. 7 and 8). The starting pore pressure of 20 MPa was varied by ± 5 MPa (PGR6-RI-01- 08) and ± 10 MPa (PGR6-RI-01-09), corresponding to a change of around 500–1000 m in the water column in a geothermal well. In total, six cycles of pore pressure increase and decrease were performed for each experiment.
The permeability was reduced when the pore pressure was reduced
International Journal of Rock Mechanics and Mining Sciences 134 (2020) 104393
(effective pressure increase), while the permeability increased for increasing pore pressure (effective pressure decrease). The total varia-tions in permeability for the pressure changes of ± 5 MPa (PGR6-RI-01- 08, Fig. 7b) ranged from 8.53*10-17 m2 to 2.57*10-16 m2, while
permeability varied from 3.84*10-16 m2 to 6.0*10-17 m2 for pressure
variations of ± 10 MPa (PGR6-RI-01-09, Fig. 8b). Permeability was slightly increased from 1.10*10-16 m2 to 1.58*10-16 m2 after six pressure
cycles for the variations of 5 MPa (PGR6-RI-01-08). Similar, a minor permanent increase from 8.65*10-17 m2 to 1.17*10-16 m2 was observed
for the effective pressure variations of 10 MPa (PGR6-RI-01-09). Comparing the mechanical dilation evolution (Figs. 7c and 8c), both samples showed a similar behaviour, which is opening during the pore pressure increase (effective pressure decrease) and closure during the pore pressure decrease (effective pressure increase). Doubling the pore pressure change led to a doubling in magnitude of the dilation change. During the pore pressure changes, the load cylinder was maintained at a constant vertical position after the maximum displacement was
reached. This way, the shear stress evolution was obtained from the measured vertical load (equation (7), Figs. 7a and 8a). When the pore pressure is varied, the effective normal stress is also changed, which can lead to slip events when a failure criterion is exceeded, resulting in a change in shear stress. For the 5 MPa effective pressure variation, a reduction in shear stress (“slip event”) was observed in the 5th and 6th cycle, reducing the shear stress by about 5 MPa. For the 10 MPa effective pressure variation the decrease in shear stress occurred in the 2nd cycle already with a larger magnitude of reduction of about 10 MPa. The vertical stress in the system could only be released towards the bottom of the sample, where the core of the sample was mounted on springs (top load piston was fixed). Since the dilation change suggested only small changes in lateral direction (Figs. 7c and 8c), we assume that this stress release resulted in a downward movement of the inner cylinder, such as a “slip event”, i.e. a small vertical displacement along the microfault. The friction coefficient could not be calculated since the normal stress was possibly disturbed by the sample and fracture geometry and needed
Fig. 4. The change in inflow and outflow rate and axial load vs. the axial displaced (a), mechanical and volume-based aperture (b), fracture volume (c) and sample
permeability (d).
to be determined numerically. The slip events resulted in a small reduction of permeability for a change of 5 MPa in pore pressure, accompanied by further fracture closure. Conversely, the slip event for the 10 MPa pore pressure change resulted in a slight increase in permeability and a dilation increase. However, the mechanical dilation was only measured at the centre of the intact portion (LIP), such that
opening or closing during slip depends on the orientation of the fracture at this location.
The fracture stiffness, χ, was obtained at each loading cycle during
pore pressure reduction (equation (8)). It was observed that the fracture stiffness increases with each loading cycle and that fractures stiffness was higher when the initial effective pressure level was higher (Fig. 9), while the permeability was not significantly changing with an increasing number of pressure cycles. Generally, the permeability changes and fracture stiffness suggested a mechanically and hydraulically sustain-able fracture.
3.4. Fracture geometry from x-ray micro-CT scan analysis
Computed x-ray CT scanning was performed using the entire sample after testing to visualize the deformation within the sample. We used a nanotom® ultra-high resolution nanoCT system with a 360◦ scanning
projection, providing a resolution for our sample size (50 by 50 mm) of about 25 μm. Computed x-ray CT scanning images of the microfaults in samples PRG6-RI-01-08 and PGR6-RI-01-09 revealed a complex micro-fault zone connecting the upper and lower notch of the sample (Fig. 10). The major fault zone was characterised by step-over structures, clearly displaced fracture faces with a visible self-propping effect, fracture branching, as well as particles or breccia distributed across the shear zone. Radial fractures and horizontal fractures were secondary features that formed during the unloading to atmospheric pressure at the end of the test and therefore do not contribute to permeability. They usually terminate against fracture branches of the central shear zone and are found in areas where the two fracture faces are in contact, thus
Fig. 5. The change in inflow and outflow rate and axial load vs. the axial displaced (a), mechanical and volume-based aperture (b), fracture volume (c), sample
International Journal of Rock Mechanics and Mining Sciences 134 (2020) 104393
minimising permeability.
True apertures were measured at defined distances of about 0.5–0.8 mm along the fracture in x-y oriented cross sections at different height levels (21 levels with about 2300 measured apertures, N, for each sample) between the bottom and top notch (Fig. 10). Closed sections (apertures smaller 0.01 mm), as well as open sections were measured in order to obtain a contact-area ratio, Rc, which was defined as the amount
of contact points (zero aperture) over the total number of measurements,
N. Based on the aperture distribution, mean aperture, contact area ratio
and the structural features of specific areas along the vertical fracture, three specific zones were identified (zone I, II and III).
For sample PGR6-RI-01-08, zone I showed mostly apertures between 0.05 and 0.2 mm with mean apertures indicated in Fig. 10. The top part of the fracture was highly damaged and crushed. Below, a zone of transpressional features was visible with a mean aperture of 0.087 and a contact area ratio of 20% indicated as zone I. Zone II was characterised by almost vertically oriented fractures (simple shear zone) showing the highest contact area ratio of about 34% and a mean aperture of 0.089
Fig. 6. Illustration of the fracture propagation process during the different stages: a) linear elastic phase, b) fracture propagation, c) shear displacement along a
principal shear plane.
Fig. 7. Shear stress (a), permeability (b) and mechanical aperture (c) evolution during effective pressure changes of 5 MPa (sample PGR6-RI-01-08). C. Kluge et al.
mm. The largest zone was zone III characterised by an inclined fracture orientation and consequent transtension showing a highest mean aper-ture of 0.17 mm and a contact-area ratio of 31%.
Sample PGR6-RI-01-09 showed a similar distribution of zone I (transpression), zone II (simple shear) and zone III (transtension). However, the mean apertures were overall higher (0.175 mm), with zone II showing the smallest mean aperture of 0.110 mm and highest
contact area-ratio of 41%. Furthermore, zone II was largest, meaning that the largest portion of the fracture was oriented vertically with simple shear features, and only small portions at the bottom show transtensional features.
Comparing samples PGR6-RI-01-08 and PGR6-RI-01-09, the latter had an overall lower mean aperture and higher contact-area ratio with the largest portion showing simple shearing, while the other had a larger
Fig. 8. Shear stress (a), permeability (b) and mechanical aperture (c) evolution during effective pressure changes of 10 MPa (sample PGR6-RI-01-09).
Fig. 9. Relative fracture closure vs. the logarithmic effective pressure change during the six effective pressure cycles (only pore pressure reduction) for sample PGR6-
International Journal of Rock Mechanics and Mining Sciences 134 (2020) 104393
zone of transtension with overall higher mean apertures and smaller contact area ratios. For both, shear zones cutting through feldspar minerals resulted in apertures of mostly zero while the largest open zones were found in quartz dominated areas. Fracture breccias or other fine and detached particles were found in every sample. The fact that they were located mostly in the bottom part of the fracture at the inflow zone indicates, that the particles were distributed and transported along the fracture, causing a further reduction in permeability.
Comparing the mechanical dilation (equation (5)) and volume-based dilation (equation (6.3)) with the aperture measured from the x-ray CT scan images, the post-testing CT images gave the largest mean aperture ranging from 0.09–0.18 mm while the mechanical dilation measured in- situ with extensometers gave 0.08–0.1 mm after failure.
4. Discussion
4.1. Microfault generation
Shear fracture geometries generated in conventional fracture permeability experiments sometimes do not represent the complex ge-ometry of single fault zone on a laboratory scale. In this section, the fracture properties, the fracture geometry regarding structural features, as well as implications of the aperture distribution for the overall permeability and strength, obtained by the PTS test, will be discussed.
Our results suggest a sufficient reproducibility for this testing
procedure in terms of mode II fracture toughness, yield point, as well as permeability and dilation evolution. Monitoring the changes in inflow and outflow volume of the pore fluid during failure allows for a good control of the fracturing process and reveals similar processes as described by Backers,37 where the fracture initiated from the outer bottom notch and propagated towards the inner top notch. However, it is not clear from the experiments whether wing fractures, as described by Backers,37 cause the initial increase in inflow rate and that the actual
shear fracture might form later and more rapidly at the point of failure. The fracture toughness values of about 15.3–16.4 MPa*m0.5 were
slightly higher compared to other granites, for example ¨Avr¨o, Aue or Mizunami granite with values between 10.9–11.5 MPa*m0.5, tested
under dry conditions.35 While these experiments were performed at
non-hydrostatic conditions, our experiments were performed with a confining stress acting on the top and bottom annulus of the sample. Furthermore, we were able to produce a single fracture at high effective pressures, which has been proven to be difficult in the past when applying mode II loading conditions in, for example, shear-box experiments.11
We favour the term “microfault” for the structures produced in our experiments, although Crider47 suggested differentiating between shear
fractures (experimentally produced) and faults (field structures). The reason is, that most of the existing experimental studies applied the term “shear fracture” to describe tensile fractures with offset, which is not adequate. Although there is a discussion about whether fracture
Fig. 10. Computed x-ray CT images and the reconstruction of the fracture system of sample PGR6-RI-01-08 (a, b) and sample PGR6-RI-01-09 (c, d). Table 3
Experimental results.
Sample ID Peff [MPa] vprop [mm/s] KIIc [MPa*m0.5] kFOI [− ] emech [mm]
PGR6-RI-01-07 20 0.1 16.42 ~520 –
PGR6-RI-01-08 20 0.3 15.71 ~230 0.07 (0.09–0.1)a
PGR6-RI-01-09 20 0.3 15.34 ~230 0.08 (0.07–0.1)a
aDuring pore pressure cycles, P
eff: effective pressure, vprop: fracture propagation velocity, KIIc: mode II fracture toughness, kFOI: permeability fold of increase, emech:
mechanical dilation after failure.
initiation in PTS experiments is in shear mode,34 the ongoing
displace-ment of up to 1.2 mm clearly generated a principle shear plane with common features found in larger scale faults.48,49 The microfault
ge-ometry of the PTS test is rather complex and can be subdivided into three domains that are also found in natural fault systems: transpression, simple shear zones, transtension.50,51 The location of these systems
along the fracture are assumed to control local permeability variation and depend on the fracture orientation, as well as on the degree of bending of the fault zone between top and bottom notch, and can also occur very locally. Transpression occurred mostly at the top of the sample, while transtension was found in the middle and lower section of the fracture zone. The fault zone structure is highly dependent on the stress and displacement magnitude, protolith and fluid flow.49 A fine
grained fault core (zone of reduced aperture), a fracture dominated damage zone and damage of linking damage structures49 or “dilational
jogs”48 were found in both our samples (Fig. 10). Detailed thin section
analysis and a comparison with other rock types might provide more insights into the development of a fault zone, microfracture density or fault zone dimensions using the PTS test, to be able to compare this to general shear zone developments.52 In addition, the permeability can be
adequately quantified due to the flow parallel orientation of fracture, meaning that compared to conventional tri-axial testing geometries, the fluid is not required to flow through a matrix before entering the frac-ture. Computed x-ray CT images also did show additional fractures such as radial fractures present in the matrix (Fig. 10). Due to the displace-ment and the interlocking of asperities the inner cylinder was pushing
against the outer cylinder, which resulted in brittle radial fracturing exclusively during unloading to atmospheric pressure at the end of the experiment (secondary feature). There was no indication of radial fracturing during displacement or pore pressure cycling since they would result in a larger permeability increase. Monitoring the electric conductivity of a silver paste applied to the circumference of the sample suggests an intact sample wall during fracture generation. To reduce the possibility of radial fracture generation, we recommend to increase the diameter of the sample while keeping the diameter of the notches. Furthermore, the top of the inner diameter needs support (steel ring) for hydrostatic pressure conditions, as hydrostatic conditions reduce the difference of compressional and tensile forces at the bottom notch, which are required to initiate shear fracture growth.37 This possibly also
explains the slightly higher load, or more precise, higher fracture toughness values, required for fracture generation.
4.2. Permeability and dilation evolution during fracturing and shearing
Quantifying and understanding the physical properties and processes that control the hydraulic-mechanical properties of fractured rocks are crucial for any engineering application in the subsurface. Experimental data of the generation and displacement of a single shear fracture zone under in-situ reservoir and fracture parallel flow conditions while measuring permeability are limited in the present literature. In the following, we will discuss our results of the permeability of microfaults framed in the context of existing approaches and outcomes in existing
Fig. 11. Normalized stress-strain curves (a), permeability evolution (b) and volume changes (c) during failure for the Odenwald granite and the experiments with
International Journal of Rock Mechanics and Mining Sciences 134 (2020) 104393
literature (Fig. 11).
In order to compare the measured permeability pre-, during and post-failure, the measured value should represent a pseudo steady-state value. To ensure these conditions we calculated the hydraulic diffusivity and the additional fluxes caused by the fracture formation. The diffu-sivity of 4.5*10-6 m2/s was calculated according to Nicholas et al.,43
resulting in a pore pressure diffusion to take less than 10 min. Furthermore, the generated fracture volume was measured to be maximum 0.2 ml within 90 min, while the applied flow rates varied between 0.4 ml/min and 1 ml/min. We concluded that both effects could be neglected for the permeability evolution during shearing, but consider the permeability during fracturing as transient.
Our results show that the magnitude of permeability enhancement is about 2 to almost 3 orders of magnitude, from less than 10−18 m2 to
more than 10−16 m,2 for a medium grained fresh granite such as the
Odenwald granite with a sample diameter of 50 mm. This is more than the 17 to 35 times permeability increase reported by Ye & Ghassemi28 on
Sierra White Granite for their double flawed samples with a complex shear zone. However, it is similar to the experiments on Westerly granite and Cerro Cristales granodiorite by Mitchell & Faulkner20 (Fig. 11b)
using the pore pressure oscillation technique in conventional tri-axial testing at effective pressures of 10–50 MPa. Similar to their study, we assume that the damage in the surrounding rock matrix is negligible and that permeability is exclusively controlled by the generated microfault. The fracture volume evolution in the experiments by Mitchell and
Faulker20 shows a similar inflection at about 70–90% of the peak stress,
although permeability increases already at lower strain, i.e. at about 60% of peak strain. The differences in fracture volume (Fig. 11c) arise from the sample and fracture geometry, or more precise, the fact that pore fluid can hardly travel through the rock matrix at the bottom and top of the sample in a tri-axial test, while there is an instant hydraulic connection and flow-parallel conditions in the PTS test. Generally, both results demonstrate that measuring permeability over more than three orders of magnitude experimentally during fracture initiation and propagation remains technically difficult. Still, similar to the experi-ments by De Paola et al.,16 the permeability evolution can be subdivided into three stages: compaction, dilation and post-failure, with the tran-sition from compaction to dilation between 60 to 80% of peak strain. Assuming that Mitchell and Faulkner20 rather measured pre-failure permeability evolution and that our results show the permeability in-crease during failure and at post-failure state, the total inin-crease by brittle faulting can exceed 3 orders of magnitude (Fig. 10).
The bulk permeability of a fault is controlled by the size of the fault zone and the aperture resulting from shear dilation. The mechanical and volume based dilation, as well as fracture volume, are commonly used to determine the aperture changes in fractures due to changes in load,5 but
present only bulk measurements for certain locations along the fracture. The x-ray CT images (Fig. 11) showed that local permeability is controlled by the distribution of aperture, which is again controlled by the fracture geometry and grain size. Similar to the study by Frash
Fig. 12. Comparison of the permeability decay curves over time for different granites and fracture types reported from literature27,57 (a). Permeability reduction due
to increasing effective pressure of non-displaced5,11,30 and displaced tensile fractures5,11,29 from literature and the microfaults from our study (b). C. Kluge et al.
et al.,12 aperture and therefore permeability are highly controlled by the
presence of en-echelon or step-over structures, as well as fault particle transportation along the fault. Experimental studies with planar struc-tures might simplify the processes involved,25–27 but do not reflect the
evolution of permeability in active faults.
We demonstrated the ability to shear a fracture at an effective pressure of 20 MPa using the PTS test. Permeability increased by a factor of 1.1 to 4.0 by shearing under in-situ conditions right after producing the shear zone. Such an increase is comparable to experiments with sheared tensile fractures in other studies.21,24 A high dependence on the
grain size is involved in the mechanical processes during shearing of a microfault, due to the sample size on a laboratory scale. The applied displacement did not exceed the grain size of the material and therefore no stick-slip behaviour, commonly observed in large scale fault mech-anisms, was observed in our experiments. The realistic microfault ge-ometry causes a heterogeneous aperture distribution, which highlights the importance of structural features, such as en-echelon or step-over structures, in faults during the displacement. We therefore also sup-port the assumption that local pressure gradients play an critical role in controlling local permeability perturbations, when fluid is directly injected into a fault zone,53 as well as when fluid is injected in an
adjacent well.
4.3. Sustainability of permeability
The sustainability of the permeability of fault structures, in other words, the resistance of a fracture towards closing caused by pressure changes, is crucial for several underground applications. In the following chapter, we will discuss our results of longevity for microfault permeability and compare those with data from tensile and displaced tensile fractures published in literature (Fig. 12a and b).
Fault compaction due to elevated normal stress can significantly reduce the permeability of a microfault. Because of the relatively short testing period, we assume that no chemical rock-fluid interactions impact the hydraulic properties of the fracture. However, we could not consider possible particle redistribution affecting permeability. We therefore recommend to analyse the effluent in future studies. The observed reduction in permeability over time after shear fracture for-mation by a factor of 2.5 to 4 might be related to three possible mech-anisms: (a) chemical rock-fluid interactions, i.e. dissolution or precipitation of minerals, (b) fines migration leading to a blockage of fluid pathways within the fault, and (c) mechanical creep, i.e. compac-tion or rearrangement of fault gouge particles. Mechanical back-slip is not possible, since the axial displacement piston was held in place after reaching a maximum of 1.2 mm.
Rock-fluid interactions, re-crystallizations or cementations are un-likely to influence permeability in our short-term experiments (several hours). Such processes evolve over long time-scales of several weeks33
and require saline solution as the pore fluid while we used distilled water, such it is unlikely that this impacted permeability. For longer term experiments, such processes should be considered to impact fault permeability. Possible fines migration cannot be assessed in this study since it requires a microstructural analysis of the microfault. Comparing our data to literature data, the permeability decay curves over time are similar to those in literature5,27 for exponential or power law decay,
although the normal effective stress range conditions (from 3 to 60 MPa) and fracture type (smooth saw-cut or displaced tensile type) differ to our study (Fig. 12a). Other authors suggest a linear decay54 of permeability
over several days, which did not fit our data. This might be caused by an insufficient resolution of permeability of the first 24 h in these tests. The main factor controlling the permeability decay cannot be explained by the data we obtained, but permeability losses up to one order of magnitude are possible within the first 16 h.
Considering pore pressure changes of ± 5 MPa to ± 10 MPa that simulate injection and production procedures in geothermal systems, permeability increased with decreasing effective pressure and decreased
by less than one order of magnitude. This process was rather reversible. In Fig. 12b we compared our permeability during the effective pressure loading to fracture permeability data from literature for dis-placed and non-disdis-placed tensile fractures. To be able to compare our sample permeability data, we need to calculate the fracture perme-ability, kf, assuming that the permeability relates to the area of flow
following Hofmann et al.5:
Asks=Amkm+Afkf (9.1)
where k is permeability, A is area of flow and the subscripts s, m, f relate to sample, matrix and fracture, respectively. The area of the sample is approximated by As =πr2 and the fracture area is calculated from the
circumference of the fracture multiplied by the aperture Af =
ahπ(2rID)2, with 2rID =r. If we now assume that the matrix permeability of the granite is negligible (km =0), we can rearrange equation (9.1) for the fracture permeability, which is given by the cubic law (kf =a2/12): a2/12 =(k
sπr2
)/
(rπah) (9.2)
This provides an equation for the hydraulic aperture similar to Hofmann et al.,5 but for a circular fracture:
ah= ̅̅̅̅̅̅̅̅̅̅̅ 12ksr 3 √ (9.3) Inserting the hydraulic aperture, ah, into the cubic law we obtain the
fracture permeability (kf =a2/12), that we can now compare to the
fracture permeability from the literature. The permeability reference value for normalisation at 10 MPa for our data is 1.97*10-12 m2.
Comparing the permeability loss during the first loading cycle to other studies5,11,29,30 the permeability loss at the respective effective pressures are similar in magnitude with displaced fractures, during the first loading cycle when normalizing permeability to the permeability at 10 MPa effective pressure (Fig. 12b). Permeability loss at lower effective pressure is strongly affected by the experimental set up and is usually higher, such that those pressures are not considered here. Generally, displaced fractures and the microfaults in our study have a higher resistance to fracture closure compared to fractures without displace-ment. Higher displacements result in a higher sustainability of fracture permeability when comparing our results to those by Watanabe et al.11
and Chen et al.29 This implies, that the type of fracturing (matched tensile, displaced tensile or shear fracturing) is a key aspect when assessing the performance and sustainability of, for example, a geothermal reservoir. Moreover, all literature values presented here5,11,29,30 result from increasing confining pressure while in our
ex-periments the pore pressure was decreased at constant confining pres-sure. In this context, we recommend to consider effective pressure coefficients31,55 instead of using Terzaghi’s effective pressure. However,
these coefficients are highly dependent on the rock type used and need to be determined experimentally.31 Changes in pore pressure, as in our
experiments, are likely to have a smaller impact on fracture perme-ability, meaning that the change in permeability per step effective pressure is larger, assuming a coefficient of 0.9 for granite.56
Consid-ering several pressure cycles, the permanent permeability change in the presented experiments with six pressure cycles was an increase by a factor less than 1.4. Other studies suggested a reduction in permeability of up to one order of magnitude after two cycles, with the highest permeability damage during the first loading stage.5 We argue that the
permeability of a microfault generated under in-situ conditions might be reversible due to the lack of plastic energy in the system. In experiments with displaced tensile fractures5 for example, the fracture preparation, i.
e. placing two sample halves together by hand, causes a high amount of plastic energy. Under external load asperities break and the fracture re-arranges itself, causing a permanent permeability decrease in every loading cycle. This is not the case for in-situ fracture generation in the PTS test. Furthermore, aperture closure analysis suggests a stiffening of the fracture zone, i.e. less closure at similar stress,32 with an increasing
